What does the research show about fasting insulin and homa-ir?

Fasting insulin and HOMA-IR

Fasting insulin is the insulin in blood after fasting, and HOMA-IR is an index calculated from it and fasting glucose. US diabetes bodies have not endorsed a diagnostic range for either. Published reference intervals top out between 11 and 18 µIU/mL, and assays can differ twofold. Higher HOMA-IR goes with more heart disease and deaths, and fasting insulin less clearly.

Checked 2026-11-13

Status: A blood test ordered by a clinician and also sold direct to consumers by some US laboratories. FDA classifies immunoreactive insulin test systems as class I devices exempt from premarket notification [1]. A joint laboratory guideline of the American Association for Clinical Chemistry and American Diabetes Association advises against routine insulin testing for most people with diabetes or at risk of it, and states that no standardized method exists to measure it [2]. An American Diabetes Association workgroup wrote in 2010 that there are no criteria by which a person could be classified as insulin sensitive or resistant [3].
Measures: The concentration of insulin in serum or plasma after fasting [1]. HOMA-IR is not measured. It is an index calculated from fasting insulin and fasting glucose, as insulin in µU/mL times glucose in mmol/L divided by 22.5 [4][5].
Also called: fasting insulin, insulin, serum insulin, insulin fasting, immunoreactive insulin, IRI, HOMA-IR, HOMA, HOMA1-IR, HOMA2-IR, homeostasis model assessment, homeostatic model assessment of insulin resistance, insulin resistance score
Tested for: insulin resistance, prediabetes screening, metabolic health, weight gain, polycystic ovary syndrome, diabetes risk, heart disease risk, low blood sugar workup

Strong several reviews agree · Moderate one review or consistent trials · Limited one or a few small studies · Untested no study in people, listed with what it rests on

Linked in research2

ModerateLinked to more new high blood pressure, 63% more in the highest against the lowest fasting insulin group across 10 cohorts, and not linked to stroke.
ModerateLinked to higher fasting insulin in women with polycystic ovary syndrome than in women without it, in both overweight and non-overweight groups, in one meta-analysis.

Reported risks0

No risks sourced yet.

Reference ranges

Each range is the one its source published, in µIU/mL. Where sources differ, each is shown.

Reference values in healthy Iranian adults, men, Tehran Lipid and Glucose Study, 2014

Non-obese healthy men aged 24 to 83, measured by electrochemiluminescence immunoassay · [6]

Range	What the source calls it
below 1.61 µIU/mL	No name given
1.61 to 11.37 µIU/mL	"reference values"
above 11.37 µIU/mL	No name given

A reference interval from one study, not a diagnostic range. No US diabetes body has endorsed a diagnostic range for fasting insulin: a joint laboratory guideline advises against routine testing, and an American Diabetes Association workgroup states that no criteria exist for classing a person as insulin sensitive or resistant [2][3]. The interval is the central 95% of results in 124 men selected as healthy and non-obese, so 1 in 20 such men falls outside it. Mean insulin fell with age. Results from another assay may not be comparable. The same study gives a reference interval for HOMA-IR of 0.63 to 2.68, an index with no unit that is not placed on this scale [6].

Reference values in healthy Iranian adults, women, Tehran Lipid and Glucose Study, 2014

Non-obese healthy women aged 24 to 83, measured by electrochemiluminescence immunoassay · [6]

Range	What the source calls it
below 2.34 µIU/mL	No name given
2.34 to 11.98 µIU/mL	"reference values"
above 11.98 µIU/mL	No name given

A reference interval from one study, not a diagnostic range, and no US diabetes body has endorsed one [2][3]. The interval is the central 95% of results in 185 women selected as healthy and non-obese. For men and women together the study gives 2.11 to 12.49 µU/mL, which its authors round to 2 to 12 [6].

Reference interval from a Brazilian laboratory database, 2024

Adults aged 20 to 60, both sexes, measured by electrochemiluminescence immunoassay · [7]

Range	What the source calls it
below 2.52 µIU/mL	No name given
2.52 to 13.14 µIU/mL	"reference interval"
above 13.14 µIU/mL	No name given

A reference interval from one study, not a diagnostic range, and no US diabetes body has endorsed one [2][3]. It was estimated by a statistical method from the stored results of 21,684 people in Rio de Janeiro after exclusions, 86% of them women, not from people examined and found healthy. The authors found sex-specific intervals not justified. The same study proposes a HOMA-IR reference interval of 0.39 to 2.86, which is not placed on this scale [7].

Reference interval in 75-year-olds, Sweden, 2013

Men and women aged 75 without diabetes · [8]

Range	What the source calls it
below 1.74 µIU/mL	No name given
1.74 to 18.27 µIU/mL	"reference interval"
above 18.27 µIU/mL	No name given

A reference interval from one study, not a diagnostic range, and no US diabetes body has endorsed one [2][3]. It is the 2.5th to 97.5th percentiles of 698 people, reported in mIU/L, which is the same number as µIU/mL. Insulin rose strongly with body mass index, and for the 574 people with a body mass index of 30 or below the interval was 1.66 to 15.05 [8].

Definition used in a US national survey analysis, 2025

US adults aged 20 and older without diabetes, excluding pregnant women, as a research definition · [5]

Range	What the source calls it
10 or below µIU/mL	No name given
above 10 µIU/mL	"hyperinsulinemia"

A definition chosen by the authors of one study to track trends, not a diagnostic range, and no US diabetes body has endorsed one [2][3]. The authors note that 10 µU/mL was above the median in every survey cycle. By this definition 41.4% of US adults without diabetes had hyperinsulinemia in 2017 to 2018, up from 28.2% in 1999 to 2000. The survey changed insulin assays over the years, and the authors adjusted results to one method. The same study defined insulin resistance as HOMA-IR above 2.6, its 66.7th percentile [5].

What has been shown to move it

Shown to lower it11

Keto dietStrongStudied in women

Weight and insulin resistance in PCOS

Shown in pooled studies to lower weight and insulin resistance in women with PCOS. Against other diets, the reviews disagreed on weight.

Compared with other diets [21]; low-calorie diets [22]; before the diet, in the same women [21][22][23]

The numbers and sources

Effect: Against other diets, ketogenic diets lowered weight 4.98 kg more (95% CI 0.91 to 9.05), BMI 1.65 kg/m2 more and insulin resistance more (MD -1.71, 95% CI -2.98 to -0.43, I² 90%). From before to after the diet, weight fell 10.77 kg and BMI 3.38 kg/m2 [21]. In 2 RCTs against low-calorie diets, body weight did not differ significantly, while BMI, fat mass, glucose and HOMA-IR fell more on the ketogenic diets [22]. A third review of 7 studies found weight, BMI, glucose, insulin, HOMA-IR, LDL cholesterol and triglycerides fell from before to after the diet [23].
Population: Women with PCOS, most with BMI above 25, in 15 studies, 10 pooled [21]; 10 studies including 3 RCTs, 4 cohorts and 2 case series [22]; 7 studies [23].
Sources: [21], [22], [23]
Caveats: Heterogeneity was high for most outcomes [21]. Before-and-after results cannot separate the diet from weight loss or time. A published letter disputed whether the definition of real-world exposure in [21] matched its inclusion criteria [24].

Alpha-lipoic acidModerate

Blood sugar

Mixed results: one large review found lower fasting glucose and HbA1c, while a review of overweight adults and neuropathy trials found no change.

The numbers and sources

Effect: Across 63 trials, fasting glucose fell 5.28 mg/dL (95% CI -7.21 to -3.35), HbA1c 0.40% (95% CI -0.66 to -0.14), fasting insulin 1.70 mU/mL and HOMA-IR 0.74 against control [8]. In 11 trials of 704 overweight or obese adults, fasting glucose (SMD 0.13, 95% CI -0.16 to 0.41) and HOMA-IR (SMD -0.23, 95% CI -0.60 to 0.15) did not differ from control [9]. In diabetic neuropathy trials, HbA1c did not change [5].
Population: 63 randomized trials [8]; adults with a BMI of 25 or more [9]; people with diabetic neuropathy [5].
Sources: [8], [9], [5]
Caveats: The reviews disagree. The overweight review rated 8 of its 11 trials at low risk of bias [9].

ChromiumModerateStudied in women

Insulin resistance

Shown to lower HOMA-IR and fasting insulin slightly in overweight adults and women with PCOS. Fasting glucose and HbA1c did not change significantly.

In 10 RCTs in 683 women with polycystic ovary syndrome, chromium lowered fasting insulin (p 0.01) and raised the insulin sensitivity index QUICKI (p 0.02) against placebo.

The numbers and sources

Effect: In 20 trials in overweight and obese adults, HOMA-IR fell 0.26 (95% CI -0.48 to -0.03) and fasting insulin 12.55 pmol/L (95% CI -23.62 to -1.47). Fasting glucose (-1.60 mg/dL, 95% CI -3.28 to 0.07) and HbA1c (-0.05 points, 95% CI -0.19 to 0.07) did not change significantly [6]. In 10 RCTs in 683 women with polycystic ovary syndrome, chromium lowered fasting insulin (p 0.01) and raised the insulin sensitivity index QUICKI (p 0.02) against placebo. Against metformin, HOMA-IR fell more on chromium (p < 0.00001) [7].
Population: Overweight and obese adults, with trial sizes of 1 to 137 [6]; women with polycystic ovary syndrome [7].
Sources: [6], [7]
Caveats: The two reviews studied different populations. The PCOS abstract gives p-values without effect sizes [7]. The overweight analysis abstract calls the fasting glucose fall significant though its interval crosses zero [6]. In 2005 FDA ranked the evidence for chromium picolinate and insulin resistance at its lowest level for a qualified claim [5].

Dark chocolateModerate

Blood vessel function and insulin resistance

Shown to improve blood vessel widening and lower insulin resistance in short trials of chocolate or cocoa, in one meta-analysis rated low to moderate quality.

Compared with control conditions the abstract does not describe [3]

HOMA-IR fell 0.67 (95% CI 0.36 to 0.98) through lower serum insulin.

The numbers and sources

Effect: Flow-mediated dilation rose 1.34% (95% CI 1.00 to 1.68) with regular intake and 3.19% after a single dose. HOMA-IR fell 0.67 (95% CI 0.36 to 0.98) through lower serum insulin. Changes in LDL cholesterol (-0.07 mmol/L, 95% CI -0.13 to 0.00) and HDL cholesterol were of borderline significance [3].
Population: 1,297 participants in 42 trials of up to 18 weeks [3].
Sources: [3]
Caveats: The authors rated the evidence low to moderate quality and called for larger, longer, independently funded trials [3].

Nordic dietModerate

Blood sugar and insulin

Mixed results: fasting glucose did not change in adults, while fasting insulin and insulin resistance fell slightly.

Compared with an average Nordic diet with the same calories [4]; control diets the source does not describe [7]

The numbers and sources

Effect: In 6 RCTs (618 participants, 6 to 26 weeks), fasting glucose did not change (-0.05 mmol/L, 95% CI -0.13 to 0.01), while fasting insulin fell 1.12 mU/L and HOMA-IR 0.34 against control diets. The insulin result depended on one study [7]. Insulin sensitivity did not change in SYSDIET [4].
Population: Adults in Nordic countries with cardiometabolic risk factors [4][7].
Sources: [4], [7]
Caveats: Two-hour glucose and the Matsuda index did not change [7].

Paleo dietModerate

Blood sugar and insulin

Mixed results: two reviews found lower fasting insulin, while a review against Mediterranean, diabetes and guideline diets found no difference in glucose, insulin or HbA1c.

Compared with control diets the source does not describe [3]; a control group the source does not describe [5]; the Mediterranean diet, a diabetes diet and a diet following Dutch Health Council advice [6]

The numbers and sources

Effect: Against Mediterranean, diabetes and Dutch guideline diets, fasting glucose (SMD -0.34, 95% CI -0.87 to 0.18), fasting insulin, HOMA-IR and HbA1c (SMD -0.38, 95% CI -0.87 to 0.11) did not differ [6]. In people with metabolic disorders, HOMA-IR fell 0.39 (95% CI 0.08 to 0.70) and fasting insulin 12.17 μU/mL (95% CI 0.08 to 24.26) [5]. In 19 RCTs, fasting insulin fell 1.01 (95% CI 0.57 to 1.45) [3].
Population: 98 people with altered glucose metabolism in 4 studies [6]; adults with metabolic disorders [5]; adults in 19 RCTs [3].
Sources: [3], [5], [6]
Caveats: The null review was small [6]. The reviews used different comparison diets.

Physical activityModerateStudied in women

Insulin resistance and weight in PCOS, in trials

Shown in trials to lower fasting insulin and insulin resistance slightly in women with PCOS. Fasting glucose did not change, at very low to low quality.

Compared with usual care or control groups [27]; before exercise, including trials with no control group [28]

The numbers and sources

Effect: Against control, exercise lowered fasting insulin (MD -2.44 µIU/mL, 95% CI -4.24 to -0.64) and HOMA-IR (MD -0.57, 95% CI -0.99 to -0.14), both very low quality, and LDL cholesterol and triglycerides. Waist fell 2.62 cm and BMI 1.02 kg/m2 after the programs. Fasting glucose and blood pressure did not change. Exercise added to diet did not differ from diet alone [27]. Vigorous exercise was linked to a 36.2% fall in HOMA-IR, a 4.2% fall in waist circumference and a 24.2% rise in VO2peak [28].
Population: 18 trials in women with PCOS [27]; 777 women in 19 studies, including uncontrolled trials, of 6 to 26 weeks [28].
Sources: [27], [28]
Caveats: Many effects were modest with wide intervals and shifted when single trials were added or removed [27]. [28] pooled controlled and uncontrolled studies.

Vitamin DModerateStudied in women

Blood sugar and insulin in PCOS

Shown in one review to lower fasting glucose and insulin slightly against placebo in women with PCOS. An earlier review found changes mainly with other nutrients.

Compared with placebo [71][72]

The numbers and sources

Effect: Across 13 RCTs (691 women), vitamin D lowered fasting glucose by 2.91 mg/dL (95% CI 1.04 to 4.78) and insulin by 1.98 µIU/mL (95% CI 0.64 to 3.32) more than placebo, with lower triglycerides, total cholesterol and LDL cholesterol [71]. An earlier review of 11 RCTs (601 women) found lower fasting glucose and HOMA-IR when vitamin D was given with other micronutrients, and lower HOMA-IR at doses under 4,000 IU a day [72].
Population: Women aged 18 and older with PCOS in 13 RCTs [71]; 601 women with PCOS in 11 RCTs [72].
Sources: [71], [72]
Caveats: The changes are small. The earlier review's abstract gives no effect size and does not report a significant effect for vitamin D alone at all doses [72].

YogaModerateStudied in women

Cycles, androgens and blood sugar in PCOS

Shown in one small meta-analysis to lower menstrual irregularity, signs of excess androgens, fasting glucose and insulin resistance in women with PCOS, at low strength of evidence.

Compared with controls the source abstract does not describe [21]; conventional physical activity, routine care or no intervention [22]

The numbers and sources

Effect: Pooling 2 RCTs, yoga lowered menstrual irregularity (MD -0.41, 95% CI -0.74 to -0.08), clinical hyperandrogenism (MD -0.70, 95% CI -1.15 to -0.26), fasting glucose (MD -0.22 mmol/L, 95% CI -0.44 to -0.01), fasting insulin (MD -28.21 pmol/L, 95% CI -43.79 to -12.63) and HOMA-IR (MD -0.86, 95% CI -1.29 to -0.43) [21]. A 2026 review of trials with 258 participants found fasting glucose fell 4.50 mg/dL more than in controls in one trial (95% CI 2.39 to 6.61) and did not differ significantly in another. In one trial, testosterone did not change significantly and LH fell [22].
Population: 515 women with PCOS in 11 studies, 2 RCTs pooled [21]; 258 women with PCOS in RCTs [22].
Sources: [21], [22]
Caveats: The authors of [21] rated the strength of evidence low. [22] could not pool its trials, and three of its five reports came from one trial.

ZincModerate

Blood sugar

Shown to lower fasting glucose about 14 mg/dL and HbA1c about 0.55% in people with and without diabetes.

Across 18 RCTs (1,023 participants) with diabetes, gestational diabetes or prediabetes, zinc lowered serum insulin (MD -2.50, 95% CI -4.69 to -0.31) and HOMA-IR (MD -1.10, 95% CI -2.05 to -0.15), and LDL cholesterol rose slightly (MD 3.46, 95% CI 1.48 to 5.43).

The numbers and sources

Effect: Across 32 placebo-controlled interventions (1,700 participants, 14 countries), zinc lowered fasting glucose by 14.15 mg/dL (95% CI -17.36 to -10.93), 2-hour glucose by 36.85 mg/dL, HbA1c by 0.55% (95% CI -0.84 to -0.27) and hs-CRP by 1.31 mg/L [3]. Across 18 RCTs (1,023 participants) with diabetes, gestational diabetes or prediabetes, zinc lowered serum insulin (MD -2.50, 95% CI -4.69 to -0.31) and HOMA-IR (MD -1.10, 95% CI -2.05 to -0.15), and LDL cholesterol rose slightly (MD 3.46, 95% CI 1.48 to 5.43) [4].
Population: People with and without diabetes [3]; people with diabetes or prediabetes [4].
Sources: [3], [4]
Caveats: The fasting glucose effect depended on diabetes status and on the zinc formulation [3].

Fiber supplementsLimited

Blood sugar

Limited evidence: psyllium lowered fasting glucose and HbA1c in type 2 diabetes but not in normal blood sugar. Other fiber results were small or inconsistent.

Across 12 RCTs in overweight and obese adults, isolated soluble fiber lowered fasting glucose by 0.17 mmol/L (95% CI -0.28 to -0.06) and fasting insulin by 15.88 pmol/L.

The numbers and sources

Effect: In people treated for type 2 diabetes, psyllium taken before meals lowered fasting glucose by 37.0 mg/dL and HbA1c by 0.97%. There was no glucose lowering in people with normal blood sugar and a modest effect in prediabetes [5]. Across 12 RCTs in overweight and obese adults, isolated soluble fiber lowered fasting glucose by 0.17 mmol/L (95% CI -0.28 to -0.06) and fasting insulin by 15.88 pmol/L [6]. A 2026 meta-analysis of 11 glucomannan trials in adults with insulin resistance-related conditions found fasting glucose 7.02 mg/dL lower (95% CI -13.56 to -0.49), an effect that lost significance under an alternative correlation assumption [7]. In a 4-week trial of 131 adults, inulin lowered 1-hour and 2-hour glucose on a glucose tolerance test in overweight or obese participants but not in healthy-weight ones, and fructooligosaccharides did not change glucose in either group [8].
Population: 35 psyllium RCTs across three groups [5]; 609 overweight and obese adults [6]; 11 glucomannan trials [7]; 131 adults in China [8].
Sources: [5], [6], [7], [8]
Caveats: Reference [5] searched clinical records held by Procter & Gamble, a psyllium supplement maker, as well as published studies, and its abstract does not report confidence intervals. The glucomannan result was not robust to the analysis method [7]. In the inulin trial, fasting insulin rose [8].

Tested, and not shown to change it12

Apple cider vinegarStrong

Blood sugar

Shown to lower fasting blood sugar and HbA1c in adults with type 2 diabetes. In apparently healthy participants, fasting glucose rose instead.

Across 9 studies in adults, ACV lowered fasting glucose by 7.97 mg/dL (95% CI -13.74 to -2.21) and HbA1c by 0.50 (95% CI -0.90 to -0.09), with no change in fasting insulin or HOMA-IR.

The numbers and sources

Effect: In 7 controlled trials in type 2 diabetes, ACV lowered fasting blood sugar by 21.9 mg/dL (95% CI -29.19 to -14.67) and HbA1c by 1.53 (95% CI -2.65 to -0.41), with no change in HOMA-IR [1]. Across 9 studies in adults, ACV lowered fasting glucose by 7.97 mg/dL (95% CI -13.74 to -2.21) and HbA1c by 0.50 (95% CI -0.90 to -0.09), with no change in fasting insulin or HOMA-IR [2]. A review of 6 studies in 317 people with type 2 diabetes found lower fasting glucose and HbA1c with vinegar [3].
Population: Adults with type 2 diabetes [1][3]. Adults with and without diabetes [2].
Sources: [1], [2], [3]
Caveats: Heterogeneity for HbA1c was high (I² 91%) [2]. Vinegar content varied across studies and samples were small [3]. In apparently healthy participants, fasting glucose rose [2].

TurmericStrong

Blood sugar

Shown to lower fasting glucose about 6 to 10 mg/dL and HbA1c about 0.3 points in adults with prediabetes or type 2 diabetes.

With piperine added, 15 RCTs (1,020 participants) showed fasting glucose 5.89 mg/dL lower, while HbA1c, fasting insulin and HOMA-IR did not differ significantly.

The numbers and sources

Effect: Across 34 RCTs in adults with prediabetes or type 2 diabetes, curcumin or turmeric lowered fasting glucose by 10.15 mg/dL (95% CI -12.59 to -7.72), HbA1c by 0.32% (95% CI -0.43 to -0.21) and HOMA-IR by 0.46 against control [6]. A meta-analysis of 63 trials drawn from 13 earlier reviews (3,706 participants) found fasting glucose 6.30 mg/dL lower (95% CI -9.33 to -3.27) and HbA1c 0.31% lower (95% CI -0.57 to -0.05) [7]. With piperine added, 15 RCTs (1,020 participants) showed fasting glucose 5.89 mg/dL lower, while HbA1c, fasting insulin and HOMA-IR did not differ significantly [8].
Population: Adults with prediabetes or type 2 diabetes [6][7]; adults, many overweight with metabolic disorders [8].
Sources: [6], [7], [8], [9]
Caveats: Heterogeneity was significant [6] and high [7]. GRADE certainty was low for fasting glucose and HbA1c in the piperine review [8]. In a network meta-analysis of 24 RCTs in prediabetes, turmeric extract was linked to fewer new cases of type 2 diabetes (RR 0.05, 95% CI 0.00 to 0.78), with low certainty and short trials [9].

Vegetarian dietStrong

Blood sugar in type 2 diabetes

Shown to lower HbA1c about 0.3 to 0.4 points against comparison diets in adults with type 2 diabetes.

Compared with comparator diets the source does not describe [1]; control diets the source does not describe [2]

In 9 RCTs (664 participants), HbA1c fell 0.29 points (95% CI -0.45 to -0.12) and fasting glucose 0.56 mmol/L, at moderate certainty, with no change in fasting insulin.

The numbers and sources

Effect: In 6 controlled trials (255 participants), vegetarian diets lowered HbA1c by 0.39 points (95% CI -0.62 to -0.15), and fasting glucose did not change significantly [1]. In 9 RCTs (664 participants), HbA1c fell 0.29 points (95% CI -0.45 to -0.12) and fasting glucose 0.56 mmol/L, at moderate certainty, with no change in fasting insulin [2].
Population: Adults with type 2 diabetes, mean age 42.5 in [1]; adults with diabetes in trials of 3 weeks or more [2].
Sources: [1], [2]
Caveats: Both reviews pooled lacto-ovo and vegan diets. Trials were small and short.

GinsengModerate

Blood sugar

Mixed results: one review found fasting glucose about 0.3 mmol/L lower, and a larger review found no significant effect on blood sugar.

Fasting insulin, HbA1c and HOMA-IR did not change overall.

The numbers and sources

Effect: Across 16 trials of 30 days or longer, ginseng lowered fasting blood glucose by 0.31 mmol/L (95% CI -0.59 to -0.03). Fasting insulin, HbA1c and HOMA-IR did not change overall [5]. A 2025 meta-analysis of 70 studies (4,506 participants) on cardiovascular risk factors found no significant effect on the glycemic profile, with a dose-response association for fasting glucose and fasting insulin [6].
Population: 770 people with and without diabetes [5]. 4,506 participants in 70 studies [6].
Sources: [5], [6]
Caveats: The two reviews disagree on fasting glucose. Two thirds of trials in [5] ran under 12 weeks, and most participants already had well-controlled glucose [5].

Green tea extractModerate

Blood sugar

Shown to lower fasting glucose, HbA1c and insulin resistance modestly with EGCG supplements in adults, with HbA1c down about 0.18 points.

HbA1c fell 0.18 percentage points (95% CI -0.35 to -0.02), and fasting insulin did not change significantly (95% CI -1.46 to 0.47).

The numbers and sources

Effect: In a 2025 meta-analysis of 41 RCTs, EGCG supplements lowered fasting glucose, HbA1c and HOMA-IR by modest amounts. HbA1c fell 0.18 percentage points (95% CI -0.35 to -0.02), and fasting insulin did not change significantly (95% CI -1.46 to 0.47) [6].
Population: Adults in 41 RCTs of EGCG [6].
Sources: [6]
Caveats: The authors describe the HbA1c and insulin effects as modest and possibly not clinically meaningful for most people [6].

Milk thistleModerate

Blood sugar in type 2 diabetes

Shown to lower fasting glucose about 27 mg/dL and HbA1c about 1 point in adults with type 2 diabetes, in low quality trials.

Across 6 RCTs (673 participants), HOMA-IR fell by 2.29 (95% CI -4.55 to -0.03) with no change in fasting insulin.

The numbers and sources

Effect: In 5 trials (270 patients), silymarin lowered fasting blood glucose by 26.86 mg/dL (95% CI -35.42 to -18.30) and HbA1c by 1.07 points (95% CI -1.73 to -0.40). Lipids did not change [9]. A 2025 network meta-analysis of 17 studies (1,337 people with type 2 diabetes) found silymarin lowered fasting glucose, HOMA-IR and HbA1c against placebo (P < 0.05) [10]. Across 6 RCTs (673 participants), HOMA-IR fell by 2.29 (95% CI -4.55 to -0.03) with no change in fasting insulin. The fall was seen in type 2 diabetes but not in fatty liver disease [11].
Population: Adults with type 2 diabetes [9][10]; adults with type 2 diabetes, alcoholic cirrhosis or fatty liver disease in Iran, Egypt and Italy [11].
Sources: [9], [10], [11]
Caveats: The authors of [9] rated the evidence as low quality, with high heterogeneity. The network meta-analysis abstract reports p-values for silymarin against placebo without effect sizes [10]. The authors of [11] called the evidence limited.

MK-677Grey marketModerateHow often is not known

Raised blood sugar and lower insulin sensitivity

Shown to raise fasting blood sugar about 5 mg/dL and lower insulin sensitivity in older adults, with poorer glucose handling in obese men.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Fasting blood glucose rose an average of 0.3 mmol/L (5 mg/dL) on MK-677, and insulin sensitivity fell [1]. In obese men, fasting glucose and insulin did not change, but an oral glucose tolerance test showed impaired glucose handling at 2 and 8 weeks [2].
Population: 65 older adults [1]; 24 obese men [2].
Sources: [1], [2]
Caveats: Two small trials.

NutsModerate

Blood sugar in type 2 diabetes

Shown to lower HbA1c by 0.07 percentage points and fasting glucose slightly in people with type 2 diabetes, in one meta-analysis of short trials.

Compared with diets with the same calories without tree nuts [12]

Fasting insulin and HOMA-IR did not change significantly.

The numbers and sources

Effect: A median 56 g a day of tree nuts lowered HbA1c (MD -0.07%, 95% CI -0.10 to -0.03) and fasting glucose (MD -0.15 mmol/L, 95% CI -0.27 to -0.02). Fasting insulin and HOMA-IR did not change significantly [12].
Population: 450 people with diabetes in 12 randomized trials of 3 weeks or more [12].
Sources: [12]
Caveats: The reviewers describe most trials as short and of poor quality [12].

Red light therapyModerate

Weight and waist size in obesity

Shown to reduce waist size, weight and BMI in people with obesity in one meta-analysis of 11 trials. Body fat percentage did not change.

Fat mass percentage (-0.28, 95% CI -1.25 to 0.69), waist-to-hip ratio and insulin did not differ.

The numbers and sources

Effect: Across 11 RCTs (569 patients), photobiomodulation improved waistline (mean difference -7.28, 95% CI -9.97 to -5.67), weight (-3.54, 95% CI -5.97 to -1.11) and BMI (-1.18, 95% CI -1.93 to -0.43), and lowered C-reactive protein, total cholesterol and HOMA-IR. Fat mass percentage (-0.28, 95% CI -1.25 to 0.69), waist-to-hip ratio and insulin did not differ [14].
Population: 569 people with obesity in 11 RCTs with varied comparators, including exercise, diet and sham light [14].
Sources: [14]
Caveats: The abstract does not give units for the mean differences. Comparators varied, and the measure of body fat itself did not change. The authors call for larger multi-center trials with longer follow-up [14].

Whole grainsModerate

Blood sugar in trials

Shown to lower the rise in blood sugar after a meal against refined foods. Longer trials found no change in fasting glucose, and one review found HbA1c fell.

Compared with similar refined foods [12]; refined grains or placebo [9]

The numbers and sources

Effect: In 14 single-meal trials, whole grain foods reduced the 2-hour glucose response by 29.71 mmol·min/L (95% CI -43.57 to -15.85) and the insulin response. In 16 trials lasting weeks to months, fasting glucose, fasting insulin and HOMA-IR did not change significantly [12]. Another review found whole grains lowered HbA1c (SMD -0.33, 95% CI -0.61 to -0.04) [9].
Population: Healthy adults in 30 trials [12]. 25 trials of adults with or without cardiovascular risk factors [9].
Sources: [12], [9]
Caveats: Most longer trials enrolled healthy people with normal blood sugar [12].

Black seed oilLimitedStudied in women

Inflammation markers

One small trial found 1,000 mg a day lowered IL-6 in 60 women with PCOS. Trials of other inflammation markers conflicted.

Fasting insulin and HOMA-IR did not differ.

The numbers and sources

Effect: In 60 women with polycystic ovary syndrome, 1,000 mg a day of oil for 12 weeks lowered serum IL-6 against placebo (adjusted mean difference -4.21, 95% CI -6.34 to -2.08) and fasting glucose by 8.16 mg/dL. Fasting insulin and HOMA-IR did not differ [9]. A review of 4 human trials found conflicting results for the adhesion molecules VCAM-1 and ICAM-1 [10].
Population: 60 women with PCOS [9]. Patients with coronary artery disease, type 2 diabetes, Hashimoto's thyroiditis or cardiovascular risk factors [10].
Sources: [9], [10]
Caveats: One small trial for IL-6 [9]. The adhesion molecule review also included 15 animal and in vitro studies, which are not counted here [10].

SermorelinLimited

Blood sugar and laboratory changes

Small studies found no change in fasting glucose, insulin or lipids in children over 1 year and in older men over 6 weeks.

In older men over 6 weeks, glucose, insulin and the glucose tolerance test did not change, lipids did not change, and no significant adverse effects were reported.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Over 1 year in children, no adverse changes in general biochemical or hormonal tests were reported, with no change in fasting glucose and no excessive rise in IGF-1 [1]. In older men over 6 weeks, glucose, insulin and the glucose tolerance test did not change, lipids did not change, and no significant adverse effects were reported [4].
Population: 110 children [1]; 11 older men [4].
Sources: [1], [4]
Caveats: Small studies without placebo groups.

Type in your own result

How it compares with the reference method

Mixed results2

Moderate

HOMA-IR against the insulin clamp

Mixed results: HOMA-IR correlated with the clamp at about 0.60 across 22 studies, and its estimates shifted twofold with the insulin assay used.

The numbers and sources

Compared with: Hyperinsulinemic euglycemic clamp
Effect: In the study that introduced it, the HOMA estimate of insulin resistance correlated with the clamp at 0.88 and had a coefficient of variation of 31%, which its authors said limits its use [4]. In a meta-analysis, log HOMA-IR correlated with the clamp at -0.60 (95% CI -0.66 to -0.53) across 22 studies [18]. In 2,321 people studied with the clamp, a decision rule built on HOMA-IR and body mass index identified insulin resistance with a sensitivity of 84.9% and a specificity of 78.7% [19]. Run on the same samples with 11 insulin assays, HOMA estimates of insulin resistance ranged from 0.8 to 2.0 in people with normal glucose, and insulin read 15% lower in heparinized plasma than in serum [20].
Population: 22 studies reporting correlations with the clamp [18]. 2,321 people, 2,138 of them without diabetes, at 17 European sites, San Antonio and the Pima Indian reservation [19]. 12 people with normal glucose, 18 with prediabetes and 67 with type 2 diabetes [20].
Sources: [4], [18], [19], [20]
Caveats: The clamp is a research procedure, and the sources state no single threshold for agreement. A US diabetes workgroup states that measures of insulin sensitivity are comparable only within individual studies [3].

Limited

Published HOMA-IR cut points

Limited evidence: published HOMA-IR cut points for people without diabetes run from 1.85 to 4.65, each set in a different population by a different method.

The numbers and sources

Compared with: Each study's own yardstick: the clamp, a population percentile or the metabolic syndrome
Effect: An Italian population study took 2.77, the lower limit of the top fifth of HOMA-IR in non-obese adults with no metabolic disorders [21]. A decision rule fitted to clamp results in Europe and the US used HOMA-IR above 4.65, or above 3.60 together with a body mass index above 27.5, noting that these corresponded to fasting insulin of 20.7 and 16.3 µU/mL [19]. A Spanish population study found 3.46 at the 90th percentile. In men without diabetes, the cut point fitted to the metabolic syndrome was 1.85 by one definition of the syndrome (ATP III) and 2.05 by another (IDF) [22]. A US survey analysis used above 2.6, the 66.7th percentile [5]. Two reference interval studies put the upper limit at 2.68 and 2.86 [6][7].
Population: 888 adults aged 40 to 79 in Bruneck, Italy, with the cut point set in 225 of them [21]. 2,321 people studied with the clamp [19]. 2,459 Spanish adults aged 20 to 92 [22]. 17,310 US adults without diabetes [5].
Sources: [21], [19], [22], [5], [6], [7]
Caveats: None of these is a diagnostic range endorsed by a US diabetes body [2][3]. A cut point set as a percentile depends on the population it came from, and all depend on the insulin assay. The Spanish study found that in women without diabetes the fitted cut point changed with age [22]. HOMA-IR is an index with no unit, so these figures are not placed on the insulin scale above.

Did not match the reference method3

Moderate

Agreement between insulin assays

Shown to differ between assays: results for the same samples varied by a factor of 2 across 11 commercial insulin assays.

The numbers and sources

Compared with: An isotope dilution mass spectrometry reference procedure, and the mean of all assays tested
Effect: Across 11 commercial assays run on 150 serum samples, insulin values varied by a factor of 2, which the authors relate to specificity, calibration or conversion factors [15]. In a workgroup convened by the American Diabetes Association, 4 of 10 commercial methods had 95% or more of results within 32% of the mass spectrometry value, and for 7 of 10 methods the bias exceeded 15.5% in 36% to 100% of individual samples [16]. A 2023 laboratory guideline states that there is still no standardized method to measure serum insulin and that attempts to harmonize assays gave greatly discordant results [2].
Population: 150 serum samples from people with various degrees of glucose intolerance [15]. 39 single-donor sera measured by 10 methods from 9 manufacturers [16].
Sources: [15], [16], [2]
Caveats: The two comparisons date from 2007 and 2009, so individual assays may since have changed. A laboratory report does not always name the assay, which limits comparison of a result with an interval or cut point set on a different one.

Limited

Results in pmol/L against results in µIU/mL

Limited evidence: a 2019 commentary reports two conversion factors in use between insulin units, the more widely used one understating insulin by about 15%.

The numbers and sources

Compared with: The conversion factor the commentary identifies as correct
Effect: Conventional units (µIU/mL, also written µU/mL or mIU/L) are based on biological activity, and SI units (pmol/L) on mass. A commentary on the history of the units reports at least two accepted conversion factors in place of a single one, and states that the more widely used factor is incorrect and underreports insulin concentrations by about 15% [17]. An American Diabetes Association workgroup asked for insulin to be reported in pmol/L and equated about 12 pmol/L with about 2 µU/mL [3]. One reference interval study printed 2.52 to 13.14 µU/mL as 15.2 to 79.2 pmol/L [7].
Population: A commentary and a workgroup statement, not studies of patients [17][3].
Sources: [17], [3], [7]
Caveats: This is why this page gives no conversion factor. A result reported in pmol/L may have been converted with either factor, and the report rarely says which.

Limited

Variation from one test to the next

One study of 90 adults without diabetes found HOMA-IR varied 26.7% within a person across 10 weekly tests, driven largely by variation in insulin.

The numbers and sources

Compared with: The same person's fasting results on 10 weekly draws
Effect: The within-person variation of HOMA-IR was 26.7% (95% CI 25.5 to 28.3), which the authors attribute largely to variability in plasma insulin. They conclude that HOMA-IR is highly variable in the same individual, which limits the value of single measurements [23].
Population: 90 adults without diabetes (52 women) from five European countries, sampled weekly on 10 occasions [23].
Sources: [23]
Caveats: One study, in people without diabetes. The abstract gives the figure for HOMA-IR and not for insulin alone [23].

What studies tie it to

Linked in research2

Moderate

High blood pressure and stroke

Linked to more new high blood pressure, 63% more in the highest against the lowest fasting insulin group across 10 cohorts, and not linked to stroke.

The numbers and sources

Effect: Comparing the highest with the lowest group of fasting insulin, the pooled relative risk was 1.63 (95% CI 1.35 to 1.97) for hypertension and 1.18 (0.87 to 1.60) for stroke. Each 50 pmol/L higher fasting insulin went with a relative risk of 1.25 (1.14 to 1.36) for hypertension and 0.999 (0.99 to 1.01) for stroke [11].
Population: 36,617 people and 4,491 cases in 10 studies of hypertension, and 27,887 people and 1,550 cases in 7 studies of stroke [11].
Sources: [11]
Caveats: Observational, from one meta-analysis, so confounding by body weight and other factors cannot be ruled out.

ModerateStudied in women

Polycystic ovary syndrome

Linked to higher fasting insulin in women with polycystic ovary syndrome than in women without it, in both overweight and non-overweight groups, in one meta-analysis.

The numbers and sources

Effect: In 8 studies, fasting insulin was higher in women with PCOS than in women without it (standardized mean difference 0.62, 95% CI 0.36 to 0.88). Compared within weight groups, it was higher in overweight women with PCOS than in overweight women without it (0.65, 0.19 to 1.12, 4 studies), and in non-overweight women with PCOS than in non-overweight women without it (0.47, 0.11 to 0.83, 5 studies). Study means ran from 13.58 to 28.66 µIU/mL in women with PCOS and from 8.00 to 25.6 µIU/mL in women without it [24]. In 28 studies using the insulin clamp, insulin sensitivity was 27% lower in women with PCOS than in controls (99% CI ±6%), independent of body mass index, and a higher body mass index lowered it a further 15% in PCOS [25].
Population: 12 studies with 450 participants in all, 8 of them on fasting insulin [24]. 28 clamp studies of premenopausal women with PCOS and controls [25].
Sources: [24], [25]
Caveats: Observational, from small studies, and the overweight comparison rests on 4 of them [24]. Fasting insulin rises with body mass index [8], which is why the comparisons within weight groups matter. The clamp is a research procedure, not a fasting insulin test [25]. Insulin assays differ (see Agreement between insulin assays).

Mixed results3

Strong

Coronary heart disease

Mixed results: two meta-analyses found no clear link between fasting insulin and coronary heart disease and one found 50% more, while HOMA-IR was linked in the one that pooled it.

The numbers and sources

Effect: In 14 Western prospective studies, the odds ratio for coronary heart disease in the top against the bottom third of fasting insulin was 1.12 (95% CI 0.98 to 1.28), and the authors concluded the association is likely to be more modest than previously suspected [9]. In a meta-analysis of adults without diabetes, the relative risk of coronary heart disease for high against low values was 1.12 (95% CI 0.92 to 1.37) for fasting insulin, 1.52 (1.31 to 1.76) for fasting glucose and 1.64 (1.35 to 2.00) for HOMA-IR. Per standard deviation it was 1.04 (0.96 to 1.12) for insulin and 1.46 (1.26 to 1.69) for HOMA-IR [10]. A third meta-analysis found a relative risk of 1.50 (95% CI 1.28 to 1.77) for the highest against the lowest fasting insulin group, and 1.16 (1.10 to 1.22) per 50 pmol/L [11].
Population: 2,649 coronary heart disease cases in 14 studies of fasting insulin [9]. 516,325 participants in 65 studies of glucose, insulin or HOMA-IR [10]. 22,379 people and 1,986 cases in 9 studies [11].
Sources: [9], [10], [11]
Caveats: Observational. Each study compared its own high and low groups, measured with its own insulin assay, so the pooled groups share no common cut point.

Limited

Developing type 2 diabetes

Limited evidence: in one cohort HOMA-IR predicted type 2 diabetes and fasting insulin alone did not clearly, and both predicted it less strongly than fasting glucose.

The numbers and sources

Effect: In a Dutch cohort, the odds ratio for type 2 diabetes in the highest against the lowest fifth was 1.9 (95% CI 0.9 to 3.6) for fasting insulin and 2.8 (1.4 to 5.6) for HOMA-IR, against 9.0 (4.4 to 18.5) for fasting glucose and 3.8 (2.0 to 7.2) for HbA1c [12]. In a US clinic sample tested for suspected prediabetes or diabetes, the odds of having prediabetes rose across quartiles of fasting insulin, to 5.04 (95% CI 3.12 to 8.12) in the highest quartile (mean 25.9 µIU/mL) against the lowest (mean 4.9 µIU/mL). The authors reported that a fasting insulin above 9.0 µIU/mL identified 80% of people with prediabetes [13].
Population: 1,349 people aged 50 to 75 without diabetes, followed a mean 6.4 years, 152 of whom developed diabetes [12]. 965 patients, mostly white, obese and female, 287 of them with prediabetes [13].
Sources: [12], [13]
Caveats: The clinic study is retrospective and cross-sectional, in people already referred for testing, and its 9.0 µIU/mL figure comes from one assay in one clinic [13]. A US diabetes workgroup states that cutoff values of insulin tied to a known size of diabetes risk have yet to be determined [3].

Moderate

Death from any cause

Mixed results: a meta-analysis of adults without diabetes tied the highest HOMA-IR group to 34% more deaths, with a weaker, borderline link for fasting insulin.

The numbers and sources

Effect: Comparing the highest with the lowest category, the pooled adjusted risk ratio for death from any cause was 1.13 (95% CI 1.00 to 1.27, P = 0.058) for fasting insulin and 1.34 (1.11 to 1.62) for HOMA-IR. For cardiovascular death it was 2.11 (1.01 to 4.41) for HOMA-IR in two studies and 1.40 (0.49 to 3.96) for fasting insulin in one [14].
Population: 26,976 adults without diabetes in 7 articles [14].
Sources: [14]
Caveats: Observational. The authors call the findings for cardiovascular death unreliable because so few studies were included [14].

HOMA-IR

Sources

US Food and Drug Administration. 21 CFR 862.1405, Immunoreactive insulin test system. Electronic Code of Federal Regulations. https://www.ecfr.gov/current/title-21/section-862.1405
Sacks DB, Arnold M, Bakris GL, et al. Guidelines and Recommendations for Laboratory Analysis in the Diagnosis and Management of Diabetes Mellitus. Clinical Chemistry. 2023. https://doi.org/10.1093/clinchem/hvad080
Staten MA, Stern MP, Miller WG, Steffes MW, Campbell SE; Insulin Standardization Workgroup. Insulin Assay Standardization: Leading to measures of insulin sensitivity and secretion for practical clinical care. Diabetes Care. 2010. https://doi.org/10.2337/dc09-1206
Matthews DR, Hosker JP, Rudenski AS, et al. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia. 1985. https://doi.org/10.1007/BF00280883
Wu C, Ke Y, Nianogo RA. Trends in Hyperinsulinemia and Insulin Resistance Among Nondiabetic US Adults, NHANES, 1999-2018. Journal of Clinical Medicine. 2025. https://doi.org/10.3390/jcm14093215
Tohidi M, Ghasemi A, Hadaegh F, et al. Age- and sex-specific reference values for fasting serum insulin levels and insulin resistance/sensitivity indices in healthy Iranian adults: Tehran Lipid and Glucose Study. Clinical Biochemistry. 2014. https://doi.org/10.1016/j.clinbiochem.2014.02.007
Schrank Y, Fontes R, Perozo AFDF, et al. Proposal for fasting insulin and HOMA-IR reference intervals based on an extensive Brazilian laboratory database. Archives of Endocrinology and Metabolism. 2024. https://doi.org/10.20945/2359-4292-2023-0483
Larsson A, Stridsberg M, Lind L. Reference values for fasting insulin in 75 year old females and males. Clinical Biochemistry. 2013. https://doi.org/10.1016/j.clinbiochem.2013.03.024
Sarwar N, Sattar N, Gudnason V, Danesh J. Circulating concentrations of insulin markers and coronary heart disease: a quantitative review of 19 Western prospective studies. European Heart Journal. 2007. https://doi.org/10.1093/eurheartj/ehm115
Gast KB, Tjeerdema N, Stijnen T, Smit JW, Dekkers OM. Insulin Resistance and Risk of Incident Cardiovascular Events in Adults without Diabetes: Meta-Analysis. PLoS ONE. 2012. https://doi.org/10.1371/journal.pone.0052036
Xun P, Wu Y, He Q, He K. Fasting insulin concentrations and incidence of hypertension, stroke, and coronary heart disease: a meta-analysis of prospective cohort studies. The American Journal of Clinical Nutrition. 2013. https://doi.org/10.3945/ajcn.113.065565
Ruijgrok C, Dekker JM, Beulens JW, et al. Size and shape of the associations of glucose, HbA1c, insulin and HOMA-IR with incident type 2 diabetes: the Hoorn Study. Diabetologia. 2018. https://doi.org/10.1007/s00125-017-4452-7
Johnson JL, Duick DS, Chui MA, Aldasouqi SA. Identifying Prediabetes Using Fasting Insulin Levels. Endocrine Practice. 2010. https://doi.org/10.4158/EP09031.OR
Zhang X, Li J, Zheng S, et al. Fasting insulin, insulin resistance, and risk of cardiovascular or all-cause mortality in non-diabetic adults: a meta-analysis. Bioscience Reports. 2017. https://doi.org/10.1042/BSR20170947
Manley SE, Stratton IM, Clark PM, Luzio SD. Comparison of 11 Human Insulin Assays: Implications for Clinical Investigation and Research. Clinical Chemistry. 2007. https://doi.org/10.1373/clinchem.2006.077784
Miller WG, Thienpont LM, Van Uytfanghe K, et al. Toward Standardization of Insulin Immunoassays. Clinical Chemistry. 2009. https://doi.org/10.1373/clinchem.2008.118380
Knopp JL, Holder-Pearson L, Chase JG. Insulin Units and Conversion Factors: A Story of Truth, Boots, and Faster Half-Truths. Journal of Diabetes Science and Technology. 2019. https://doi.org/10.1177/1932296818805074
Otten J, Ahrén B, Olsson T. Surrogate measures of insulin sensitivity vs the hyperinsulinaemic-euglycaemic clamp: a meta-analysis. Diabetologia. 2014. https://doi.org/10.1007/s00125-014-3285-x
Stern SE, Williams K, Ferrannini E, et al. Identification of Individuals With Insulin Resistance Using Routine Clinical Measurements. Diabetes. 2005. https://doi.org/10.2337/diabetes.54.2.333
Manley SE, Luzio SD, Stratton IM, Wallace TM, Clark PM. Preanalytical, Analytical, and Computational Factors Affect Homeostasis Model Assessment Estimates. Diabetes Care. 2008. https://doi.org/10.2337/dc08-0097
Bonora E, Kiechl S, Willeit J, et al. Prevalence of insulin resistance in metabolic disorders: the Bruneck Study. Diabetes. 1998. https://doi.org/10.2337/diabetes.47.10.1643
Gayoso-Diz P, Otero-González A, Rodriguez-Alvarez MX, et al. Insulin resistance (HOMA-IR) cut-off values and the metabolic syndrome in a general adult population: effect of gender and age: EPIRCE cross-sectional study. BMC Endocrine Disorders. 2013. https://doi.org/10.1186/1472-6823-13-47
Carobene A, Kilpatrick E, Bartlett WA, et al. The biological variation of insulin resistance markers: data from the European Biological Variation Study (EuBIVAS). Clinical Chemistry and Laboratory Medicine. 2025. https://doi.org/10.1515/cclm-2024-0672
Jin Y, Sun F, Yang A, et al. Insulin-like growth factor binding protein-1 and insulin in polycystic ovary syndrome: a systematic review and meta-analysis. Frontiers in Endocrinology. 2023. https://doi.org/10.3389/fendo.2023.1279717
Cassar S, Misso ML, Hopkins WG, et al. Insulin resistance in polycystic ovary syndrome: a systematic review and meta-analysis of euglycaemic-hyperinsulinaemic clamp studies. Human Reproduction. 2016. https://doi.org/10.1093/humrep/dew243

Everything in the library

Linked in research2

Reported risks0

Reference ranges

Reference values in healthy Iranian adults, men, Tehran Lipid and Glucose Study, 2014

Reference values in healthy Iranian adults, women, Tehran Lipid and Glucose Study, 2014

Reference interval from a Brazilian laboratory database, 2024

Reference interval in 75-year-olds, Sweden, 2013

Definition used in a US national survey analysis, 2025

What has been shown to move it

Shown to lower it11

Tested, and not shown to change it12

How it compares with the reference method

Mixed results2

Did not match the reference method3

What studies tie it to

Linked in research2

Mixed results3

Related

Sources